1/* $NetBSD: bufq_disksort.c,v 1.13 2016/11/16 10:42:14 pgoyette Exp $ */
2/* NetBSD: subr_disk.c,v 1.61 2004/09/25 03:30:44 thorpej Exp */
3
4/*-
5 * Copyright (c) 1996, 1997, 1999, 2000 The NetBSD Foundation, Inc.
6 * All rights reserved.
7 *
8 * This code is derived from software contributed to The NetBSD Foundation
9 * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
10 * NASA Ames Research Center.
11 *
12 * Redistribution and use in source and binary forms, with or without
13 * modification, are permitted provided that the following conditions
14 * are met:
15 * 1. Redistributions of source code must retain the above copyright
16 * notice, this list of conditions and the following disclaimer.
17 * 2. Redistributions in binary form must reproduce the above copyright
18 * notice, this list of conditions and the following disclaimer in the
19 * documentation and/or other materials provided with the distribution.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
23 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
24 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
25 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
26 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
27 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
28 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
29 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
30 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
31 * POSSIBILITY OF SUCH DAMAGE.
32 */
33
34/*
35 * Copyright (c) 1982, 1986, 1988, 1993
36 * The Regents of the University of California. All rights reserved.
37 * (c) UNIX System Laboratories, Inc.
38 * All or some portions of this file are derived from material licensed
39 * to the University of California by American Telephone and Telegraph
40 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
41 * the permission of UNIX System Laboratories, Inc.
42 *
43 * Redistribution and use in source and binary forms, with or without
44 * modification, are permitted provided that the following conditions
45 * are met:
46 * 1. Redistributions of source code must retain the above copyright
47 * notice, this list of conditions and the following disclaimer.
48 * 2. Redistributions in binary form must reproduce the above copyright
49 * notice, this list of conditions and the following disclaimer in the
50 * documentation and/or other materials provided with the distribution.
51 * 3. Neither the name of the University nor the names of its contributors
52 * may be used to endorse or promote products derived from this software
53 * without specific prior written permission.
54 *
55 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
56 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
57 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
58 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
59 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
60 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
61 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
62 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
63 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
64 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
65 * SUCH DAMAGE.
66 *
67 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94
68 */
69
70#include <sys/cdefs.h>
71__KERNEL_RCSID(0, "$NetBSD: bufq_disksort.c,v 1.13 2016/11/16 10:42:14 pgoyette Exp $");
72
73#include <sys/param.h>
74#include <sys/systm.h>
75#include <sys/buf.h>
76#include <sys/bufq.h>
77#include <sys/bufq_impl.h>
78#include <sys/kmem.h>
79#include <sys/module.h>
80
81/*
82 * Seek sort for disks.
83 *
84 * There are actually two queues, sorted in ascendening order. The first
85 * queue holds those requests which are positioned after the current block;
86 * the second holds requests which came in after their position was passed.
87 * Thus we implement a one-way scan, retracting after reaching the end of
88 * the drive to the first request on the second queue, at which time it
89 * becomes the first queue.
90 *
91 * A one-way scan is natural because of the way UNIX read-ahead blocks are
92 * allocated.
93 */
94
95struct bufq_disksort {
96 TAILQ_HEAD(, buf) bq_head; /* actual list of buffers */
97};
98
99static void bufq_disksort_init(struct bufq_state *);
100static void bufq_disksort_put(struct bufq_state *, struct buf *);
101static struct buf *bufq_disksort_get(struct bufq_state *, int);
102
103BUFQ_DEFINE(disksort, 20, bufq_disksort_init);
104
105static void
106bufq_disksort_put(struct bufq_state *bufq, struct buf *bp)
107{
108 struct bufq_disksort *disksort = bufq_private(bufq);
109 struct buf *bq, *nbq;
110 int sortby;
111
112 sortby = bufq->bq_flags & BUFQ_SORT_MASK;
113
114 bq = TAILQ_FIRST(&disksort->bq_head);
115
116 /*
117 * If the queue is empty it's easy; we just go on the end.
118 */
119 if (bq == NULL) {
120 TAILQ_INSERT_TAIL(&disksort->bq_head, bp, b_actq);
121 return;
122 }
123
124 /*
125 * If we lie before the currently active request, then we
126 * must locate the second request list and add ourselves to it.
127 */
128 if (buf_inorder(bp, bq, sortby)) {
129 while ((nbq = TAILQ_NEXT(bq, b_actq)) != NULL) {
130 /*
131 * Check for an ``inversion'' in the normally ascending
132 * block numbers, indicating the start of the second
133 * request list.
134 */
135 if (buf_inorder(nbq, bq, sortby)) {
136 /*
137 * Search the second request list for the first
138 * request at a larger block number. We go
139 * after that; if there is no such request, we
140 * go at the end.
141 */
142 do {
143 if (buf_inorder(bp, nbq, sortby))
144 goto insert;
145 bq = nbq;
146 } while ((nbq =
147 TAILQ_NEXT(bq, b_actq)) != NULL);
148 goto insert; /* after last */
149 }
150 bq = nbq;
151 }
152 /*
153 * No inversions... we will go after the last, and
154 * be the first request in the second request list.
155 */
156 goto insert;
157 }
158 /*
159 * Request is at/after the current request...
160 * sort in the first request list.
161 */
162 while ((nbq = TAILQ_NEXT(bq, b_actq)) != NULL) {
163 /*
164 * We want to go after the current request if there is an
165 * inversion after it (i.e. it is the end of the first
166 * request list), or if the next request is a larger cylinder
167 * than our request.
168 */
169 if (buf_inorder(nbq, bq, sortby) ||
170 buf_inorder(bp, nbq, sortby))
171 goto insert;
172 bq = nbq;
173 }
174 /*
175 * Neither a second list nor a larger request... we go at the end of
176 * the first list, which is the same as the end of the whole schebang.
177 */
178insert: TAILQ_INSERT_AFTER(&disksort->bq_head, bq, bp, b_actq);
179}
180
181static struct buf *
182bufq_disksort_get(struct bufq_state *bufq, int remove)
183{
184 struct bufq_disksort *disksort = bufq->bq_private;
185 struct buf *bp;
186
187 bp = TAILQ_FIRST(&disksort->bq_head);
188
189 if (bp != NULL && remove)
190 TAILQ_REMOVE(&disksort->bq_head, bp, b_actq);
191
192 return (bp);
193}
194
195static struct buf *
196bufq_disksort_cancel(struct bufq_state *bufq, struct buf *buf)
197{
198 struct bufq_disksort *disksort = bufq->bq_private;
199 struct buf *bq;
200
201 TAILQ_FOREACH(bq, &disksort->bq_head, b_actq) {
202 if (bq == buf) {
203 TAILQ_REMOVE(&disksort->bq_head, bq, b_actq);
204 return buf;
205 }
206 }
207 return NULL;
208}
209
210static void
211bufq_disksort_fini(struct bufq_state *bufq)
212{
213
214 KASSERT(bufq->bq_private != NULL);
215 kmem_free(bufq->bq_private, sizeof(struct bufq_disksort));
216}
217
218static void
219bufq_disksort_init(struct bufq_state *bufq)
220{
221 struct bufq_disksort *disksort;
222
223 disksort = kmem_zalloc(sizeof(*disksort), KM_SLEEP);
224 bufq->bq_private = disksort;
225 bufq->bq_get = bufq_disksort_get;
226 bufq->bq_put = bufq_disksort_put;
227 bufq->bq_cancel = bufq_disksort_cancel;
228 bufq->bq_fini = bufq_disksort_fini;
229 TAILQ_INIT(&disksort->bq_head);
230}
231
232MODULE(MODULE_CLASS_BUFQ, bufq_disksort, NULL);
233
234static int
235bufq_disksort_modcmd(modcmd_t cmd, void *opaque)
236{
237
238 switch (cmd) {
239 case MODULE_CMD_INIT:
240 return bufq_register(&bufq_strat_disksort);
241 case MODULE_CMD_FINI:
242 return bufq_unregister(&bufq_strat_disksort);
243 default:
244 return ENOTTY;
245 }
246}
247